If you need k percentile values then I think you could do it in O(n log(k)) with introselect. You most definitely don't need O(n log(n)).
> Then again, maybe there are algorithms letting you compute percentiles without sorting?
I think you may be able to do so probabilistically, which is often good enough? Additionally, one observation about percentiles is that you don't need to sort all of the data; you can just hold on to the top N results in something like a Heap.
Extracting percentile information doesn't require sorting. Even if you're too lazy to implement anything custom, a few invocations of std::nth_element will likely outperform sorting: https://en.cppreference.com/w/cpp/algorithm/nth_element
Yeah, you can histogram the values if you just need an approximate answer (or if you have few enough discrete values, you can just have a bin per value). Once you have an approximate answer, if you need a more precise answer, you can go through the data again and only consider values within the right bin (which hopefully will be a lot fewer, if you've chosen your bins wisely) and then use something like std::nth_element.
Then again, maybe there are algorithms letting you compute percentiles without sorting?